Oxidation of levulinic acid for the production of maleic anhydride: breathing new life into biochemicals

Article abstract of DOI:10.1039/c5gc01000d

Levulinic acid (LA) is a biomass-derived platform chemical that could play a central role in emerging industries as an intermediate that facilitates production of bio-based commodities. In this context, we present a novel, catalytic pathway for the synthesis of maleic anhydride (MA) via oxidative cleavage of the methyl carbon in LA over supported vanadates. The approach is demonstrated in a continuous flow, packed bed reactor, and we have observed that VO_x supported on SiO₂ achieves single-pass MA yields as high as 71%. Preliminary analysis suggests that LA might compete with butane as an industrial MA feedstock. Finally, bifunctional LA and monofunctional 2-pentanone display contrasting oxidative cleavage selectivities, indicating that methyl carbon cleavage during vapor phase oxidation over supported vanadates is unique to LA.

Full text of DOI:10.1039/c5gc01000d

Green Chemistry
View Article Online
View Journal
PAPER
Oxidation of levulinic acid for the production
of maleic anhydride: breathing new life into
Cite this: DOI: 10.1039/c5gc01000d
biochemicals†
Anargyros Chatzidimitriou and Jesse Q. Bond*
Levulinic acid (LA) is a biomass-derived platform chemical that could play a central role in emerging
industries as an intermediate that facilitates production of bio-based commodities. In this context, we
present a novel, catalytic pathway for the synthesis of maleic anhydride (MA) via oxidative cleavage of the
methyl carbon in LA over supported vanadates. The approach is demonstrated in a continuous flow,
x 2
packed bed reactor, and we have observed that VO supported on SiO achieves single-pass MA yields as
Received 11th May 2015,
Accepted 2nd July 2015
high as 71%. Preliminary analysis suggests that LA might compete with butane as an industrial MA feed-
stock. Finally, bifunctional LA and monofunctional 2-pentanone display contrasting oxidative cleavage
selectivities, indicating that methyl carbon cleavage during vapor phase oxidation over supported vana-
dates is unique to LA.
DOI: 10.1039/c5gc01000d
www.rsc.org/greenchem
oxygenated hydrocarbons are an attractive class of products.
Their functionality makes them useful intermediates in
Introduction
Concern over diminishing fossil reserves and the environ- numerous applications; thus, they have robust, established
mental impact of their utilization has expanded interest in the markets. Further, from a petroleum chemistry standpoint, they
production of bio-based commodities. In this respect, abun- are challenging targets that require oxidative functionalization
dant lignocellulose is often viewed as an ideal alternative of relatively inert feedstocks. Typically, this occurs under
source of industrial carbon. Because the transportation sector severe conditions and forms products that are more reactive
accounts for the bulk of oil consumption, lignocellulosic bio- than their precursors; hence, it is difficult to control selectivity,
fuels are commonly proposed as an avenue for mitigating oil and oxygenated hydrocarbons are expensive relative to alkane
2
–4
reliance; unfortunately, lignocellulosic carbon is locked in fuels.
In general, functionally equivalent oxygenates can be
oxygen-rich polysaccharides and lignin. In biofuel production, produced from biomass through partial deoxygenation, which
their inherent functionality is problematic because the ideal is frequently attractive compared to alkane functionalization.
target products—liquid alkanes—are devoid of heteroatoms Because oxygen imparts reactivity, one can often transform
and functional moieties. Although technologically possible, biomass under mild conditions, where judicious catalyst selec-
synthesizing energy-dense fuels from lignocellulose requires tion can control product selectivity. As a chemicals feedstock,
extensive feedstock defunctionalization, substantial hydrogen biomass thus becomes increasingly attractive relative to fossil
1
input, and (frequently) carbon–carbon bond formation.
hydrocarbons with the amount of oxygen in the target
Accordingly, advanced biofuels are cumbersome and expens- product. With these criteria in mind, we contrast petrochem-
ive, and they have failed to compete in current markets. ical production of maleic anhydride (MA) with an alternative
With the caveat that biorefineries must target production of approach based on the oxidation of levulinic acid (LA).
saleable, large-market commodities, lignocellulose is more MA is a petrochemical commodity that has a relatively high
likely to succeed as a feedstock where its inherent functional- oxygen content (O : C = 0.75), and it is challenging to manufac-
ity—instead of comprising the central challenge—offers a ture selectively from hydrocarbons. Further, it is used indust-
distinct competitive advantage. In this context, partially rially for the production of butanediol, tetrahydrofuran,
γ-butyrolactone, 2-pyrrolidone, N-methyl-2-pyrrolidone, and a
range of polymers and surfactants such that it has a well-estab-
lished market. MA can be supplied through the aerobic oxi-
Department of Biomedical and Chemical Engineering, College of Engineering &
Computer Science, Syracuse University, Syracuse, NY 13244, USA.
E-mail: jqbond@syr.edu; Tel: +(315)443-2550
5,6
dation of butane (0.30–0.70 $ per kg, 2014) or benzene
5
,7,8
(
0.80–1.60 $ per kg, 2014).
Benzene oxidation occurs over
†
Electronic supplementary information (ESI) available. See DOI: 10.1039/
c5gc01000d
supported vanadium oxides at temperatures from 623–673 K.
This journal is © The Royal Society of Chemistry 2015
Green Chem.

View Article Online
Paper
Green Chemistry
It is generally the more selective route, delivering single pass selectivity is attainable at temperatures of 573–598 K, which is
MA yields in excess of 70%. Alternatively, vanadium-phosphor- mild compared to conventional butane oxidation (663–703 K).
ous oxides (VPO) are used to produce MA through butane oxi- Further, MA has an established market and is (on a molar
dation at slightly higher temperatures (663–703 K). Under basis) roughly ten-fold more valuable than refinery alkanes,
industrial conditions, optimal MA selectivities (65–75%) are suggesting that it could be an attractive commercial target for
reported at roughly 85% butane conversion, giving maximum emerging LA producers. Before discussing the approach
single pass yields of 50–60% during butane oxidation. further, we note that two prior reports have outlined MA pro-
Although benzene oxidation is milder and slightly more selec- duction via liquid-phase oxidation of 5-hydroxymethylfurfural
3
4
2
tive, it is not sufficiently so in either respect to offset the cost (HMF) over VO(acac) (53% MA yield) and vanadium substi-
3
5
of using benzene as an MA feedstock, and butane oxidation tuted heteropolyacids (64% MA yield). Because HMF pro-
9
has become the dominant technology. Over the last decade, duces LA upon hydrolysis, these approaches might entail
MA prices have ranged from 1.50–2.00 $ per kg in a global similar chemistries to those described here; however, LA is
1
0,11
market of roughly 1500–2000 kilotons per annum (kta).
favored as an MA feedstock from a practical standpoint. HMF
Presently, MA prices are on the lower side of this range, which is difficult to prepare selectively from lignocellulose since it
likely stems from the recent expansion of oil and natural gas generally requires either a preliminary glucose–fructose iso-
7
36,37
production in the United States. Based on butane prices over merization step or the use of exotic systems,
whereas LA
the past year and reported MA yields of 70% under industrial can be prepared directly from cellulose via dilute acid hydro-
3
3
conditions of full recycle, we estimate that butane purchasing lysis. This suggests that LA will always be a lower cost com-
will contribute $0.25–$0.60 per kg to the cost of MA modity than HMF and thus a more realistic MA precursor.
production.
From an economic standpoint, BioFine has indicated that
As an alternative, we have found that levulinic acid (LA) can cellulose-derived LA can be produced for roughly $0.21 per
3
3
be employed as an MA precursor. LA is a 5-carbon keto-acid kg. Based on this price and a yield of 70%, the cost of LA
that can be synthesized from hexose- and pentose-containing would contribute roughly $0.36 per kg to MA manufacture,
polysaccharides that comprise the bulk of lignocellulose which is comparable to estimated butane expenses in the con-
(Fig. 1).
ventional approach. While BioFine’s figure is likely optimistic,
Pathways for LA production from both C and C sugars are LA production costs could increase roughly 2.5-fold to $0.53
5
6
well-described, and the underlying methodologies extend to per kg before LA is unable to compete with butane at $1.05 per
1
2–14
7
numerous biomass feedstocks.
LA itself has no substantial kg, where it peaked in 2008 and approached in 2012. Cer-
commercial demand; however, it offers intriguing chemical tainly, the LA-based approach becomes more attractive as the
functionality and can be upgraded to multiple potentially sale- price of butane increases. Further, although 70% is likely a
able products including oxygenated fuel additives such as ceiling on attainable MA yields during well-optimized butane
1
5–21
22,23
gamma-valerolactone,
levulinates,
methyltetrahydrofuran,
alkyl oxidation under recycle conditions, we see no reason that the
and best single-pass yield reported here (71%) represents an upper
2
4,25
26
27–32
and alkyl valerates; alkane fuels;
niche specialty chemicals, such as the herbicide δ-aminolevuli- limit on attainable MA selectivity during LA oxidation; there-
3
3
nic acid (DALA). First generation attempts at commercializa- fore, future optimization could make the LA-based approach
tion of LA-based refineries have largely explored production of even more attractive.
biofuels, such as ethyl levulinate; however, due to the afore-
Our preliminary assessment of process economics is sim-
mentioned challenges facing the production of lignocellulosic plistic, relying solely on a comparison of estimated feedstock
biofuels, LA has made little headway in the transportation costs for the carbon source in butane- and LA-based strategies
sector, and development of LA-based refineries remains stag- for MA production. Although it is essential that detailed
nant. We were thus intrigued to learn that simple, vapor- design studies consider additional aspects of these processes,
phase, aerobic oxidation of LA over supported vanadates deli- including reactor sizing; materials of construction; operating
vers MA at 71% of the theoretical maximum yield. Good MA temperatures and pressures; separations; and feed recycle
before concluding that an LA-based strategy could be profit-
able in the near term, initial feasibility determinations are
generally best made by assessing the economic potential of a
technology based on raw material costs and revenue generated
3
8
through product sales. For the purposes of demonstrating
that an LA-based approach could potentially be feasible along-
side present inexpensive butane, the above analysis is sufficient.
Results and discussion
Prior to summarizing trends in LA conversion and product
selectivity with variation in operating conditions and catalyst
Fig. 1 Bio-based and petrochemical strategies for the production of
maleic anhydride (MA).
Green Chem.
This journal is © The Royal Society of Chemistry 2015

View Article Online
Green Chemistry
Paper
selection, we facilitate a comprehensive discussion of selecti- trace quantities (<1%). This suggests that, in line with mechan-
vity by considering the landscape of oxidative and non- istic expectations, SA forms as the initial cleavage product but
oxidative reactions that occur in this system.
subsequently undergoes oxidative dehydrogenation (ODH) to
As shown in Fig. 2, the primary reaction leading to MA is form MA. ODH is kinetically possible over supported vana-
4
2,43
likely oxidative cleavage of the C –C bond of LA. According to dates,
and it is thermodynamically favorable under our
4
5
mechanistic insights gleaned from 2-butene and 2-butanone reaction conditions. For reference, the oxidative dehydrogena-
oxidation, this cleavage should result in formation of equi- tion of succinic anhydride to form maleic anhydride occurs
3
9,40
−1
molar quantities of succinic acid and formaldehyde.
We with ΔG ≈ −157 kJ mol at 548 K. MA should thus form ir-
note that the formation of succinic acid and formaldehyde reversibly and be the dominant product so long as LA oxidative
−
1
from LA occurs with ΔG ≈ −326 kJ mol at 548 K, and we cleavage is slow relative to SA oxidative dehydrogenation
thus expect oxidative cleavage to occur irreversibly. Succinic (otherwise, SA would be observed as a major product based on
acid can then reversibly form its anhydride by dehydration, the proposed reaction pathways). The anticipated C
1
co-
which is thermodynamically favorable over the temperature product, formaldehyde, was never observed in our experi-
−1
range considered here (ΔG ≈ −11.6 kJ mol
Because of the relatively high temperatures and low water will further oxidize to formic acid and/or carbon oxides (CO
partial pressures in our system, succinic anhydride (SA) should under the reported conditions. Here, CO were the only experi-
at 548 K). ments; however, it is reasonable to expect that formaldehyde
x
)
x
be the dominant (equilibrium) form of succinic acid at our mentally observed C products, implying that the coproduct of
1
experimental conditions. Selective formation of succinic acid/ MA formation is always lost to combustion.
succinic anhydride (70–80% yield) via vapor-phase oxidative
4 5 1 4
Just as cleavage of the C –C bond of LA leads to C and C
cleavage of LA has been reported by Dunlop in the patent lit- products, analogous cleavage of the C –C bond should result
3
4
4
1
2 3
erature, but this result has not been confirmed in the peer- in C (acetaldehyde or acetic acid) and C products (3-oxo-
reviewed literature. Our experiments have employed similar propanoic acid or malonic acid). Both acetaldehyde and acetic
catalysts and conditions to those described in Dunlop’s acid were detected; however, their formation cannot be solely
patent; however, MA was instead recovered as the dominant attributed to C
3 4
–C bond cleavage since sequential cleavage of
oxidative cleavage product whereas SA was only observed in products may also form C
C
4
2
products. Malonic acid and
Fig. 2 Pathways consuming LA and its derivatives during oxidation over supported vanadates.
This journal is © The Royal Society of Chemistry 2015
Green Chem.

View Article Online
Paper
Green Chemistry
3
-oxo-propanoic acid were never observed, but this cannot be ing in the formation of nCO
x
. In each experiment described
taken as conclusive evidence that C –C cleavage does not below, all detectable reaction products were quantified, and
3
4
occur. Malonic acid is thermally unstable and will decarboxylate carbon balances closed to within 12%.
4
4
to form CO
2
and acetic acid at 343 K; therefore, it is unlikely
2 3
γ-Al O has a relatively high surface area, it supports good
to be recovered under our conditions. Acrylic acid and propion- vanadate dispersion at high areal loadings, and the patent lit-
aldehyde were observed in trace quantities (<0.5% selectivity). erature indicates it as a preferred vanadate support for the oxi-
4
1
These species likely form through non-oxidative C–C scission dative cleavage of LA. Accordingly, our initial investigations
i.e., cracking) of levulinic acid and/or reaction intermediates. probed the influence of operating conditions on LA conversion
At high temperatures and in the presence of acid sites, LA and product selectivity over VO /γ-Al (Fig. 3).
dehydration leads to the formation of α-angelicalactone (AAL, LA conversion increases with temperature and contact time
ΔG ≈ −11 kJ mol at 548 K) and β-angelicalactone (BAL, ΔG ≈ until 598 K, where complete LA conversion is observed at all
(
x
2 3
O
−
1
10 kJ mol− at 548 K). Methylvinylketone (MVK) also forms contact times. MA selectivity is favored by increasing reaction
1
45
−
4
6
alongside CO through decarbonylation of LA. Reaction pro- temperature, which suggests that oxidative cleavage of LA pro-
ducts additionally contained C diones (DIO), specifically 1,3- ceeds with a higher barrier than competing side reactions.
5
cyclopentanedione (DIOS) and its dehydrogenated analog, At all temperatures, we further observe that MA selectivity
4
-cyclopentene-1,3-dione (DIOU). DIOS is an isomer of α and improves with contact time even under conditions of complete
β-angelicalactone, and electronic structure calculations suggest LA conversion (see 598 K, Fig. 3). This may indicate that MA
that its formation from either angelicalactone (ANG) is slightly forms as a secondary product during LA oxidation. Over VO
x
/
−
1
2 3
favorable, occurring with ΔG ≈ −1 to −2 kJ mol at 548 K. γ-Al O , the highest MA selectivities (50–60%) were observed
DIOS is thus anticipated in appreciable quantities over this under conditions of complete LA conversion at 598 K and
range of temperatures, provided that its formation is not contact times from 4–16 min. Because of the complexity of
limited by the rate of ANG isomerization. Subsequently, ODH this reaction landscape, it is difficult to assign molar selectiv-
of DIOS should result in irreversible formation of its unsatu- ities to all products as we have done for MA. Accordingly,
−
1
rated analog, DIOU (ΔG ≈ −177 kJ mol at 548 K). Our obser- Fig. 4 summarizes complete carbon distributions for the reac-
vation that DIOU is generally the dominant DIO form tion products recovered during LA oxidation over VO /γ-Al
reinforces our conclusion that oxidative dehydrogenation is Only ANG, MA, DIOU, and CO were produced in substan-
x
2 3
O .
x
facile under our experimental conditions. It is worth mention- tial quantities (>5%), and their percentages are given discretely
ing that DIOS and DIOU are both intriguing, multifunctional in Fig. 4. Note that both AAL and BAL are included in the ANG
molecules. To our knowledge, their synthesis from LA has not label. Acetic acid, acrylic acid, acetaldehyde, propionaldehyde
been previously reported, and future efforts directed toward and MVK were observed in sufficient quantities to warrant
optimizing their selectivity may establish them as a new class inclusion in the carbon balance (1–5%), and these species are
of bio-based platform intermediates. Finally, over the reported lumped into the category “other.”
temperature range, any C
n
species illustrated in Fig. 2 might
Below 573 K, angelicalactones dominate the product distri-
additionally undergo complete or partial combustion, result- bution. Their selectivity decreases slightly with increasing
x 2 3
Fig. 3 LA conversion and MA selectivity as a function of temperature and contact time over VO /γ-Al O . MA selectivity is defined as the molar
quantity of MA formed normalized by the molar quantity of LA consumed. Contact times are calculated as the molar loading of vanadium in the
reactor normalized by the molar feed rate of LA. pLA = 0.016 bar, pO2 = 0.33 bar.
Green Chem.
This journal is © The Royal Society of Chemistry 2015

View Article Online
Green Chemistry
Paper
Fig. 4 Product carbon distributions as a function of contact time. (a) 523 K, (b) 548 K, (c) 573 K, (d) 598 K.
contact time and dramatically with increasing temperature. At
Thermodynamic analysis of the landscape shown in Fig. 2
5
73 and 598 K, product selectivity shifts primarily toward MA reveals further insights into phenomena governing selectivity
and CO , and they displace ANG as the major reaction pro- in this system. Oxidative cleavage of LA to produce SA and
x
ducts. This is tentatively attributed to LA dehydration being ODH of SA to form MA are both exceedingly favorable and
(apparently) kinetically facile relative to oxidative cleavage. At should be considered irreversible. In contrast, with the excep-
low temperatures, the extent of LA oxidation is insignificant, tion of oxidative dehydrogenation of DIOS, each step leading
and the bulk of LA consumption occurs through dehydration to the formation of DIOU should occur favorably, but reversi-
to form ANG. If oxidative cleavage is a higher barrier process bly. We therefore expect that, given a sufficient residence time,
than dehydration, it is plausible that temperature increases LA, AAL, BAL, and DIOS can all ultimately be converted to
shift selectivity toward oxidative cleavage products. Further, LA either MA or DIOU. Both of these form irreversibly; thus, the
dehydration is endothermic and, while it becomes thermo- distribution of MA and DIOU in the reaction products is likely
dynamically favorable, it is not irreversible in this temperature controlled by the relative rates of two key (irreversible) reac-
−
1
range (ΔG ≈ −7 to −18 kJ mol from 523–598 K); therefore, tions: oxidative cleavage of LA and oxidative dehydrogenation
ANG yields are likely equilibrium limited, and ANG should of DIOS. Increasing reaction temperature appears to shift
revert to LA if the equilibrium position is displaced by LA con- selectivity toward MA at the expense of DIOU, suggesting that
sumption. Elevated temperatures increase the rate of irrevers- oxidative cleavage of LA proceeds with a higher barrier than
ible oxidative cleavage of LA and should therefore drive the oxidative dehydrogenation of DIOS. Finally, we observe that
system to complete conversion and shift selectivity toward MA increases in MA selectivity are always accompanied by an
and CO . Alternatively, it may be possible that direct oxidative increase in CO selectivity. We expect that the C coproduct of
x
x
1
cleavage of either ANG or DIOS leads to MA formation. This is LA oxidative cleavage should form CO
x
under our conditions;
potentially consistent with increasing MA selectivities as a thus, a commensurate increase in CO yield with MA yield is
x
function of contact time; however, the mechanism of such is anticipated. However, stoichiometry implies that LA oxidative
presently uncertain.
cleavage should result in product carbon distributed between
This journal is © The Royal Society of Chemistry 2015
Green Chem.

View Article Online
Paper
Green Chemistry
CO
on the order of 1 : 1. This suggests that multiple pathways lead increase in the rate of LA oxidative cleavage relative to the rate
to CO in this system. of DIOS dehydrogenation in response to increases in O
The influence of molecular oxygen was considered over pressure, i.e., irreversible oxidative cleavage of LA appears
x
and MA in a 1 : 4 ratio, whereas we generally observe ratios in MA selectivity is likely attributed to a disproportionate
x
2
VO /γ-Al O , and our observations are presented in Fig. 5.
higher order in O than does irreversible ODH of DIOS.
x
2
3
2
Even under oxygen lean conditions (0.010 bar O
, O
2 2
: LA =
Because the majority of steps outlined in Fig. 2 are pro-
0
.62), we observe baseline conversion of LA. This is attributed posed to be reversible, efforts to enhance the selectivity of MA
to non-oxidative pathways, primarily dehydration, such that in this system can be reduced to two goals: suppressing ir-
ANG comprised 80% of the product carbon distribution. An reversible formation of DIOU via ODH and controlling irrevers-
additional 16% of the product carbon existed as DIO, with DIOS ible formation of CO
and DIOU recovered in 5 : 3 molar ratio. It is worth noting that varying the nature of the catalyst support. Bifunctional LA is
increasing O pressure shifts the DIO distribution completely sensitive to the acid/base character of the catalyst, and such
x
. Both might plausibly be addressed by
2
toward the unsaturated product, DIOU, which is consistent with functionalities could reduce MA selectivity. For example, ANG
our interpretation that DIOU forms via ODH of DIOS. LA con- formation is promoted by acid sites; as such, an acidic support
version increases with O partial pressure (Fig. 5b), and we attri- could facilitate DIOS formation and thus drive selectivity
2
bute this to an increase in the rates of irreversible, oxidative toward DIOU. Further, the oxidative capacity of supported
reactions that allow complete consumption of LA, ANG, and vanadates generally increases with the reducibility of the
4
7
DIOS and shift selectivity toward DIOU, MA, and CO
ing O pressure initially favors DIOU formation, while further help to limit the extent of over-oxidation in this system.
increments drive formation of MA and CO . Fig. 5b illustrates a Support influence was examined by comparing the perform-
x
. Increas- support. Accordingly, decreasing support reducibility may
2
x
clear, first order dependence of MA selectivity on O
2
pressure. ance of VO
x
/γ-Al
2
O
3
to that of VO
x
/SiO
2
, VO
x
/TiO
2
, and un-
Since SA, MA, and DIOU all form irreversibly, this enhancement supported V
2 5
O
. Results are summarized in Table 1.
2
Fig. 5 Influence of O .
partial pressure on LA conversion and MA selectivity. T = 623 K, pLA = 0.0158 bar, WHSV = 0.128 min−1
Table 1 Comparison of LA oxidation over bulk and supported vanadates. DIO includes both DIOS and DIOU
Carbon distribution (%)
b
Conversion
(%)
MA yield
(%)
CB
a
Entry
Catalyst
T (K)
τ (min)
ANG
MA
CO
x
DIO
Other
MA : CO
x
(%)
1
2
3
4
5
6
7
VO
VO
VO
x
x
x
/SiO
/TiO
/Al
2
548
548
548
548
523
548
573
1.95
0.39
78
280
3.9
3.9
3.9
99
93
98
97
43
100
100
50
26
33
18
1
48
71
2
3
5
12
83
1
43
25
27
15
2
44
57
36
61
34
21
5
35
37
15
8
26
45
8
14
2
4
3
8
7
2
6
4
1.2
0.4
0.8
0.7
0.5
1.3
1.5
96
89
100
98
98
88
2
O
2
3
2
V O
5
VO
VO
VO
x
x
x
/SiO
/SiO
/SiO
2
2
2
0
100
a
b
Contact time. Carbon balance.
Green Chem.
This journal is © The Royal Society of Chemistry 2015

View Article Online
Green Chemistry
Paper
x 2 x 2
At 548 K, VO /SiO and VO /TiO are significantly more
active than VO /γ-Al O and V O (entries 1–4). Both achieve
x
2
3
2 5
complete LA conversion at contact times below 2 min, whereas
complete LA conversions were not achieved until contact times
of roughly 80 min over VO /γ-Al O and 300 min over bulk
x
2 3
2 5
V O . Comparing selectivities at complete LA conversion, we
observe that MA selectivity increases with catalyst identity in
the order of V O < VO /TiO < VO /γ-Al O < VO /SiO . Each
2
5
x
2
x
2
3
x
2
material showed comparable selectivities toward ANG (<15%),
while the remaining product distribution varied with support Fig. 6 Enol mediated oxidative cleavage of methyl ketones.
identity. Bulk V O appears relatively selective toward DIOU,
2
5
which comprises 45% of the product carbon distribution. This
suggests that bulk V facilitates DIOS dehydrogenation to a
greater extent than it does LA cleavage. DIOU selectivity acetaldehyde and acetic acid, whereas analogous cleavage of
decreases over VO /γ-Al (26%) and further over VO /SiO
the terminal enol will form formaldehyde (C ) and propanoic
15%) and VO /TiO (8%). Over bulk V , VO /γ-Al , and acid (C ). Prior studies of 2-butanone oxidation have reported
VO /TiO , carbon selectivity towards CO exceeds that of MA, overwhelming selectivity toward C products rather than C /C
2 5
O
x
2
O
3
x
2
1
(
x
2
2
O
5
x
2 3
O
3
x
2
x
2
1
3
4
0,49,50
suggesting these materials promote combustion more so than products,
oxidative cleavage. Only VO /SiO achieves an MA : CO carbon enol stability. Electronic structure calculations indicate that
ratio greater than unity, implying that it is effective at limiting the internal enol of LA is also the more stable form such that
over-oxidation. Since VO /SiO appeared to suppress both DIO one might similarly expect internal C–C cleavage during LA
and CO formation, it was further examined (entries 5–7). Con- oxidation.
sistent with results obtained over VO /γ-Al O , MA selectivity
Oxidative cleavage of LA’s methyl group (to form SA) has
increases dramatically with temperature over VO
73 K, we obtained an MA (carbon) selectivity of 57% and a condensed-phase oxidations and/or homogeneous oxidants,
MA : CO ratio of 1.54, which corresponds to a molar MA yield and there is evidence that these systems will induce terminal
suggesting that product selectivity reflects
x
2
x
x
2
x
x
2 3
52–55
x
/SiO
2
. At been previously reported;
however, most accounts describe
5
x
5
2
of 71%. It is evident that the nature of the support can impact C–C cleavage in any methyl ketone. In contrast, the only
activity and selectivity in this system; however, each support description of selective cleavage of the terminal carbon during
employed here varies in its acid/base character, and the struc- gas-phase oxidation of methyl ketones was reported in
x
ture and function of VO clusters varies strongly with support Dunlop’s patent, which describes LA oxidation over supported
4
8
41
and VO
observed variations will require a more comprehensive charac- that the unique selectivity in this system arises from the
terization study than is warranted in the present discussion.
bifunctional nature of LA. To probe this, we compared the oxi-
x
loading.
Probing fundamental sources of the vanadates. His observations and those reported here suggest
It is interesting to note that insights from the 2-butanone dative cleavage of LA to its nearest monofunctional ketone ana-
oxidation literature suggest that MA is an unanticipated logue, 2-pentanone (Table 2), and we observed comparable
4
0,49,50
product in this system.
mediated by enol formation, and enolization of any methyl
ketone—a designation that includes both LA and 2-butanone 2-pentanone oxidation primarily yields internal scission pro-
should lead to one of two possible enol forms. One will have ducts (C2 and C3 aldehydes/acids) whereas only trace quan-
a CvC bond in the terminal position, and one will have a tities of C products were recovered. In contrast, roughly 80%
Oxidative ketone cleavage is oxidative cleavage rates with each molecule.
Consistent with prior accounts of 2-butanone oxidation,
—
4
CvC bond in an internal position (Fig. 6).
In the case of 2-butanone, the internal enol is more vage products, namely MA. This implies a unique stabilization
4
of the cleavage selectivity during LA oxidation goes to C clea-
5
0,51
stable,
2
and its oxidative scission yields two C products, of terminal C–C cleavage during LA oxidation; however, we
Table 2 Comparison of oxidative cleavage rates and C
4
x 2 3
selectivities for LA and 2-pentanone on VO /γ-Al O
a
b
c
Feed rate
τ
Conversion
(%)
Cleavage rate
C
4
CB
Entry
Feed
T (K)
(μmol min−
1
)
(min)
(μmol min g−1
)
−1
(%)
(%)
1
2
3
4
5
2-Pentanone
2-Pentanone
2-Pentanone
LA
573
585
598
573
598
79
79
79
79
79
7.80
7.80
7.80
7.80
7.80
14
49
95
99
100
22
35
59
41
50
1
2
3
85
89
96
96
96
90
90
LA
a
b
Contact time. Ratio of product carbon recovered in C
1
and C
4
(terminal) cleavage products to the total quantity of carbon recovered in
c
oxidative cleavage products. Carbon balance.
This journal is © The Royal Society of Chemistry 2015
Green Chem.

View Article Online
Paper
Green Chemistry
2
x
cannot pinpoint the mechanism by which this occurs was limited to 2.7VO /nm , due to the low density and reactiv-
5
6
without further characterization of catalytic sites and oxidation ity of the surface hydroxyls. The vanadate solution was added
kinetics.
to the support and the resulting slurry was evaporated at 338 K
under continuous mixing. The catalyst was crushed and sieved
to achieve uniform particle size and finally was activated by
holding the sample under flowing air for 4 hours at 773 K.
Materials and methods
Reagents
Reactor setup
Bulk vanadium pentoxide (98%, Acros) was used with no
further preparation. γ-alumina (low soda, Strem Chemicals),
titanium oxide (Aeroxide® P25, Acros), amorphous silica (99%
Davisil grade 633, Sigma), ammonium metavanadate (99%
Sigma) and oxalic acid (98% anhydrous, Acros) were used for
catalyst synthesis. Maleic acid (98%, Fisher), acetaldehyde
All reaction experiments were run in the gas phase in a 1/2″
downflow stainless steel tubular reactor. The catalyst bed was
held halfway through the length of the tube between two
pieces of quartz wool, while the remainder of the tube was
filled with quartz chips to ensure uniform mixing and mini-
mize the dead volume of the reactor. A 12″ long ceramic
furnace (Omega) was used to heat the reactor. Prior to each
experiment, the catalytic reactor was calcined at 773 K for
(
(
99.5%, Sigma), acetic acid (99.8%, Acros), propionaldehyde
99%, Acros), propionic acid (99%, Acros), acrylic acid (99%,
Acros), 4-cyclopentene-1,3-dione (95%, Sigma Aldrich), bute-
none (99%, Sigma), α-angelica lactone (MP Biomedicals), CO
4
hours.
A preheated mixing section was held at a temperature of
23 K, and separate feeds of LA (introduced using a syringe
2 2
(99.99% Praxair) and CO (5% in N , Airgas) were used as sup-
4
plied from the manufacturers for standard preparations and
instrument calibrations. Acetone (99.9%, Fischer) was used as
a solvent for all reactor samples and standard preparations.
Levulinic acid (LA, 98%, Sigma Aldrich) and 2-pentanone
pump, Cole Parmer Series 100) and a gas stream consisting of
and He (controlled by mass flow controllers, Valco model
00), were first heated to 423 K and introduced to the mixing
stage. The flowrates of LA, O , and He were adjusted to achieve
the desired LA and O partial pressures. The entire delivery
O
2
1
2
(99%, Acros) were used as reactor feeds. Purified water used in
2
catalyst preparation and HPLC mobile phases was prepared in
house by sequential reverse osmosis, UV oxidation, and double
system was heat traced, and LA partial pressures were always
held below LA vapor pressure at the system temperature to
prevent condensation. The gaseous mixture was subsequently
heated to reaction temperature in a preheating stage and then
introduced into the reactor. All temperatures were monitored
with K-type thermocouples (Omega) and controlled using PID
controllers (Series 16A, Love Controls).
At the exit of the catalytic reactor, the gaseous effluent was
scrubbed using an acetone filled trap, which was typically
immersed in an ice bath to allow rapid cooling and capture of
ion exchange. He (99.999%, Airgas) was used as a diluent, O
99.999%, Airgas) and O (1% in He with 1% Ar, Praxair) were
2
(
2
used as oxidants, while air (Zero and Medical Grade, Airgas)
was used for calcinations both in the catalyst synthesis phase
and for the in situ calcination of the catalytic reactor prior to
reaction.
Catalyst synthesis
2 3 2 2
γ-Al O , TiO and amorphous SiO were considered for condensable reaction products. After initial introduction of LA
catalyst synthesis. Catalysts were prepared by evaporation and oxidant feeds to the catalyst bed, reactors were permitted
to dryness of a mixture consisting of the support and 48 hours to reach a steady state. Data reported here were sub-
ammonium metavanadate dissolved in aqueous, 1 M oxalic sequently collected over roughly 200 hours on stream for a
acid. The molar quantity of vanadium employed for each given catalyst sample. During this time, reactors were period-
support was calculated based on the BET surface areas of the ically returned to the initial operating conditions, and com-
oxide supports (Table 3), obtained using a commercial system parison of rates to the initial steady state data indicated that
(
Micromeritics ASAP 2020), with the aim of creating an amor- no significant activity losses occurred over the course of a
phous, two-dimensional vanadia monolayer in contact with given experiment. Though long-term stability was not
the oxide support. assessed, catalysts employed here were observed to be fully
For Al and TiO /nm was the targeted regenerable for multiple reaction cycles (≈5) via calcination in
2
2
O
3
2 x
supports, 9.5VO
coverage, while in the case of the SiO support, the coverage air at 773 K.
2
Table 3 Surface area characterization of employed catalysts
Weight
% of V
Support SBET
Catalyst SBET
Surface density 2
(V atoms per nm )
2
−1
2
−1
Entry
Catalyst
Support
n/a
(m g
)
(m g
)
1
2
3
4
V
2
O
/γ-Al
5
56
n/a
231
481
56
4
216
333
31
VO
VO
VO
x
x
x
2
O
3
Al
SiO
P25
2
O
3
15.7
9.6
4.3
9.5
2.6
9.5
/SiO
/TiO
2
2
2
Green Chem.
This journal is © The Royal Society of Chemistry 2015

View Article Online
Green Chemistry
Paper
All product species reported here, with the exception of CO
were recovered in the condenser. All condensed species, with
the exception of MA, were quantitatively analyzed by an Agilent
x
References
1 J. Q. Bond, A. A. Upadhye, H. Olcay, G. A. Tompsett, J. Jae,
R. Xing, D. M. Alonso, D. Wang, T. Y. Zhang, R. Kumar,
A. Foster, S. M. Sen, C. T. Maravelias, R. Malina,
S. R. H. Barrett, R. Lobo, C. E. Wyman, J. A. Dumesic
and G. W. Huber, Energy Environ. Sci., 2014, 7, 1500–
1523.
2 B. J. Nikolau, M. A. D. N. Perera, L. Brachova and
B. Shanks, Plant J., 2008, 54, 536–545.
3 P. N. R. Vennestrøm, C. M. Osmundsen, C. H. Christensen
and E. Taarning, Angew. Chem., Int. Ed., 2011, 50, 10502–
10509.
7
890 GC equipped with an FID detector and an HP-INNOWAX
column. MA was quantified using an Agilent series 1100 HPLC
equipped with a Hi-Plex column. For this analysis, we
employed a pH 2 aqueous H SO mobile phase and quantified
2 4
MA concentrations using a VWD detector. Because of the rela-
tively high temperatures and low water partial pressures in our
reactor, we anticipate that MA exits our reactor as its anhy-
dride; however, because of the aqueous nature of our analysis,
MA was quantified here as maleic acid. The vapor phase (CO
was sent to an in-line chromatograph (GC 7890 Agilent). CO
x
)
2
was resolved from CO using a packed column (Restek Shin-
Carbon ST Micropacked) and quantified via TCD response
relative to a helium reference.
Qualitative product identification was achieved using an
Agilent 7890 GC-MS equipped with an Agilent 5975C MS detec-
tor and an HP-INNOWAX column. Diacid and anhydride peaks
were poorly resolved in the HP-INNOWAX column. To make
qualitative product assignments and confirm the presence of
4 J. Q. Bond, D. Martin Alonso and J. A. Dumesic, in Aqueous
Pretreatment of Plant Biomass for Biological and Chemical
Conversion to Fuels and Chemicals, ed. C. E. Wyman, Wiley
Blackwell, Oxford, UK, 2013, ch. 5, pp. 61–102.
5 K. Lohbeck, H. Haferkorn, W. Fuhrmann and N. Fedtke, in
Ullmann’s Encyclopedia of Industrial Chemistry, Wiley-VCH
Verlag GmbH & Co. KGaA, 2000, DOI: 10.1002/14356007.
a16_053.
C
4
diacids and anhydrides in our reaction products, the
6 C. Delhomme, D. Weuster-Botz and F. E. Kuhn, Green
Chem., 2009, 11, 13–26.
7 Normal Butane Spot Price, Mont Belvieu, TX, The Wall
Street Journal Database, https://www.quandl.com/data/
WSJ/BUTANE-Butane-normal-Mont-Belvieu-Texas, (accessed
March 2015).
product stream was bubbled through methanol. This con-
verted diacids (succinic, maleic, and fumaric) and their anhy-
drides into their methyl and dimethyl esters, which were easily
resolved via GC-MS and could be identified from their frag-
mentation patterns (ESI†).
8
“Regional Breakdown of Benzene Value Chain Prices”, ICIS
Chemical Business, 2015, 287, 28.
9
T. R. Felthouse, J. C. Burnett, B. Horrell, M. J. Mummey
and Y. J. Kuo, Kirk-Othmer Encyclopedia of Chemical
Technology, John Wiley & Sons, Inc., 2000, DOI: 10.1002/
0471238961.1301120506051220.a01.pub2.
Conclusion
In summary, we have demonstrated that MA can be produced
in good yield and continuous operation (71%) via aerobic, oxi-
dative cleavage of LA over vanadium oxides. This route is facili- 10 “Maleic Anhydride (MA): 2010 World Market Outlook and
tated through a mechanism that is unique for bifunctional LA
Forecast Been Recently Released by MarketPublishers.
and offers an exciting link between lignocellulosic biomass
com”, Business Wire, 2010.
and large, existing commodity markets. Moreover, the inherent 11 J. Dietrich, “Maleic anhydride”, ICIS Chemical Business,
reactivity of LA may allow the design of a milder, more selec- 2014, 285, 30.
tive oxidation strategy, providing it with a competitive advan- 12 B. Girisuta, L. Janssen and H. J. Heeres, Ind. Eng. Chem.
tage over butane as an MA feedstock. In contrast to prior Res., 2007, 46, 1696–1708.
efforts geared toward biofuel production, this approach could 13 G. M. G. Maldonado, R. S. Assary, J. A. Dumesic and
offer near-term viability and expand the industrial utilization
of lignocellulosic carbon.
L. A. Curtiss, Energy Environ. Sci., 2012, 5, 8990–8997.
14 D. Martin Alonso, S. G. Wettstein, M. A. Mellmer,
E. I. Gurbuz and J. A. Dumesic, Energy Environ. Sci., 2013,
6, 76–80.
1
1
5 L. E. Manzer, Appl. Catal., A, 2004, 272, 249–256.
6 W. R. H. Wright and R. Palkovits, ChemSusChem, 2012, 5,
1657–1667.
Acknowledgements
The authors acknowledge financial support for this work from
the National Science Foundation (Award number 1454346). We 17 M. G. Al-Shaal, W. R. H. Wright and R. Palkovits, Green
additionally thank Andreas Heyden and Muhammed Mamun Chem., 2012, 14, 1260–1263.
University of South Carolina) for assistance with electronic 18 O. A. Abdelrahman, A. Heyden and J. Q. Bond, ACS Cata.,
(
structure calculations to quantify thermodynamic data for key
2014, 4, 1171–1181.
reactions in the landscape of Fig. 2. Finally, we would like to 19 O. A. Abdelrahman, H. Y. Luo, A. Heyden, Y. Román-
thank Morris Argyle (Brigham Young University) for sharing Leshkov and J. Q. Bond, J. Catal., 2015, 329, 10–21.
his insights regarding the synthesis of oxide-supported 20 I. T. Horvath, H. Mehdi, V. Fabos, L. Boda and L. T. Mika,
vanadates.
Green Chem., 2008, 10, 238–242.
This journal is © The Royal Society of Chemistry 2015
Green Chem.

View Article Online
Paper
Green Chemistry
2
2
1 I. T. Horvath, Green Chem., 2008, 10, 1024–1028.
2 M. G. Al-Shaal, A. Dzierbinski and R. Palkovits, Green
Chem., 2014, 16, 1358–1364.
39 Y. Takita, K. Nita, T. Maehara, N. Yamazoe and T. Seiyama,
J. Catal., 1977, 50, 364–372.
40 E. McCullagh, J. B. McMonagle and B. K. Hodnett, Appl.
Catal., A, 1993, 93, 203–217.
41 A. P. Dunlop and S. Smith, US Pat, US2676186A, 1954.
42 E. A. Mamedov and V. Cortés Corberán, Appl. Catal., A,
1995, 127, 1–40.
43 A. Khodakov, B. Olthof, A. T. Bell and E. Iglesia, J. Catal.,
1999, 181, 205–216.
2
3 D. Martin Alonso, J. Q. Bond and J. A. Dumesic, Green
Chem., 2010, 12, 1493–1513.
2
2
4 P. J. Fagan and L. E. Manzer, US Pat, US7153996, 2006.
5 E. I. Gürbüz, D. Martin Alonso, J. Q. Bond and
J. A. Dumesic, ChemSusChem, 2011, 4, 357–361.
6 J. P. Lange, R. Price, P. Ayoub, J. Louis, L. Petrus, L. Clarke
2
2
2
2
3
3
3
3
and H. Gosselink, Angew. Chem., Int. Ed., 2010, 49, 4479–4483. 44 C. L. Perrin, A. Flach and M. N. Manalo, J. Am. Chem. Soc.,
7 J. C. Serrano-Ruiz, D. J. Braden, R. M. West and
2012, 134, 9698–9707.
45 R. H. Leonard, Ind. Eng. Chem., 1956, 48, 1330–1341.
J. A. Dumesic, Appl. Catal., B, 2010, 100, 184–189.
8 D. Martin Alonso, J. Q. Bond, J. C. Serrano-Ruiz and 46 J. Q. Bond, D. Martin Alonso, R. M. West and
J. A. Dumesic, Green Chem., 2010, 12, 992–999. J. A. Dumesic, Langmuir, 2010, 26, 16291–16298.
9 J. C. Serrano-Ruiz, D. Wang and J. A. Dumesic, Green 47 Z. Wu, H.-S. Kim, P. C. Stair, S. Rugmini and S. D. Jackson,
Chem., 2010, 12, 574–577. J. Phys. Chem. B, 2005, 109, 2793–2800.
0 D. Martin Alonso, J. Q. Bond, D. Wang and J. A. Dumesic, 48 I. E. Wachs, Appl. Catal., A, 2011, 391, 36–42.
Top. Catal., 2010, 54, 447–457.
49 Y. Takita, K. Inokuchi, O. Kobayashi, F. Hori, N. Yamazoe
and T. Seiyama, J. Catal., 1984, 90, 232–240.
50 Y. S. Wladimir, M. Thomas and P. Helmut, Z. Phys. Chem.,
2007, 222, 129–151.
51 S. A. Nolen, C. L. Liotta, C. A. Eckert and R. Glaser, Green
Chem., 2003, 5, 663–669.
1 J. Q. Bond, D. Wang, D. Martin Alonso and J. A. Dumesic, J.
Catal., 2011, 281, 290–299.
2 D. Martin Alonso, S. G. Wettstein, J. Q. Bond, T. W. Root
and J. A. Dumesic, ChemSusChem, 2011, 4, 1078–1081.
3 J. J. Bozell, L. Moens, D. C. Elliott, Y. Wang,
G. G. Neuenscwander, S. W. Fitzpatrick, R. J. Bilski and 52 J. Liu, Z. Du, T. Lu and J. Xu, ChemSusChem, 2013, 6, 2255–
J. L. Jarnefeld, Resour. Conserv. Recycl., 2000, 28, 227–239. 2258.
4 Z. Du, J. Ma, F. Wang, J. Liu and J. Xu, Green Chem., 2011, 53 I. Podolean, V. Kuncser, N. Gheorghe, D. Macovei,
3
3
3
3
3
1
3, 554–557.
5 J. Lan, J. Lin, Z. Chen and G. Yin, ACS Catal., 2015, 5,
035–2041.
V. I. Parvulescu and S. M. Coman, Green Chem., 2013, 15,
3077–3082.
54 B. Singh and S. Sahai, J. Indian Chem. Soc., 1991, 68, 208–
2
6 V. V. Ordomsky, V. L. Sushkevich, J. C. Schouten, J. van der
Schaaf and T. A. Nijhuis, J. Catal., 2013, 300, 37–46.
7 M. Mascal and E. B. Nikitin, Angew. Chem., Int. Ed., 2008,
209.
55 D. S. van Es, F. van der klis and J. van Haveren, World
Intellectual Property Organization Pat, WO2012044168A1,
2012.
4
7, 7924–7926.
8 A. A. Upadhye, W. Qi and G. W. Huber, AIChE J., 2011, 57, 56 I. E. Wachs and B. M. Weckhuysen, Appl. Catal., A, 1997,
292–2301. 157, 67–90.
2
Green Chem.
This journal is © The Royal Society of Chemistry 2015